Continuing miniaturization of circuit components and increasing complexity of boards is driving the need for automated optical inspection. Many types of automated equipment combine cameras and computers to form a vision system that assists in the identification, location, or processing of parts. Such vision systems are also used to correctly position robotic equipment parts which call for extremely high positioning accuracy.
One of the many challenges in designing a computer vision system is the difficulty of obtaining adequate visual contrast for accurately identifying or locating a desired feature of a part being processed. The inability to locate the desired feature of the part, or the generation of inaccurate location data by the vision system may result in mere processing inconvenience or may have more costly and damaging consequences.
For example, if the vision system is designed to generate an error condition when it is unable to locate a desired feature, the error condition may trigger the automated equipment to shut down processing of the part, causing the work to be stopped and therefore inefficiency in production.
As another example, a slightly impaired ability to locate a desired feature of a part being processed could cause the vision system to generate an inaccurate location result. When this occurs, the vision system is unaware of the inaccuracy of the location data, and therefore no error code is generated. Accordingly, the automated equipment does not shut down processing of the part, but continues to operate. This may have disastrous results if any position of the processing involves robotic contact with the part.
For example, vision systems are often used to apply adhesive to silicon chips during packaging. The adhesive is applied in predetermined locations on the silicon chip by an adhesive dispensing needle. Because the silicon chips are relatively small, the adhesive must be applied in a very controlled and precise manner, and accordingly, the needle comes into very close proximity to the surface of the chip. However, the automated adhesive dispensing equipment must be accurate enough to come into the required close proximity with the chip without allowing the needle to make actual contact with the chip's surface since actual contact can result in damage to the chip.
If the adhesive is applied in the incorrect location, or if the needle used to dispense the adhesive makes contact with, and therefore damages, the chip, the silicon chip becomes unusable, thereby increasing production costs and decreasing production efficiency.
However, a more serious consequence results if, as a result of inaccurate vision system location data, the adhesive dispensing needle comes into contact with either the chip or the automated equipment plate on which the chip is mounted and causes the needle to bend. A bent needle can contribute to the scrapping of many subsequent production parts. While a subsequent silicon chip may be located correctly by the vision system, the bent needle will cause the adhesive to be dispensed in the incorrect location on all future parts until the problem is observed by the equipment operator and the needle replaced. This can be extremely costly if a bent needle occurs early in a production run of expensive integrated circuit assemblies.
Depending on the chosen manufacturing process and chemical properties of the films and substrates of a silicon die, the surface of the die may vary in color. For example, while there are a variety of processes that can produce varying colors of parts, one color-varying effect is due to the thickness of a layer of silicon nitride that is grown on the silicon wafer during processing. Different thicknesses of this layer result in different colors as the wavelengths of light interact with the thickness of the silicon nitride layer. Accordingly, one silicon die manufactured by a given process with a first thickness of a silicon nitride layer may have a light-green appearance, while another type of die manufactured using the same process but applying a silicon nitride layer of a different second thickness may have a deep-blue appearance.
Automated adhesive dispensing equipment processing these parts needs to accurately determine the location of the silicon die prior to dispensing adhesive. Typical equipment uses an array of light-emitting diodes (LEDs) to illuminate the silicon die. The most common color of light used is red, likely due to the availability, low cost, and high light intensity produced by red LEDs. This array of LEDs may be located in a ring around the camera lens, in a planar array directed onto the product through a half-silvered mirror, or other illumination configurations.
Color begins in light that contains more of some wavelengths than others, stimulating the three types (Red (R), Green (G), Blue (B)) of color receptor cones (in a human's eye) unequally. These R, G and B cone responses can be applied to describe the color sensation that results from any mixture of light wavelengths.
A camera interprets wavelengths of light to create a perception of color. The camera's interpretation of the wavelengths of light is based on the trichromatic theory of color vision by which any color may be generated by a combination of red, green, and/or blue light. This theory therefore employs additive color mixing.
FIG. 1 illustrates a coarse color wheel for light. As illustrated, colors that lie opposite one another in the color wheel are considered complementary colors. Images taken of objects of a given color appear with very low contrast when illuminated with light of a complementary color.
FIGS. 2 and 3 illustrate example black-and-white images of silicon dies 4a, 4b as would be seen by a vision system of a conventional automated fluid dispensing equipment that uses red LEDs for illumination. The actual silicon die 4a pictured in the image of FIG. 2 has a green tint and the background plate 2 is a dark grey or black color. Accordingly, since the color green is not opposite the color red on the color wheel of FIG. 1, green is not a complementary color of red. Accordingly, red illumination light results in good contrast between the die 4a and background 2. In FIG. 3, however, the actual silicon die 4b has a blue tint. In this case, the color blue is a complementary color of red, and therefore red illumination light results in low contrast. Red illumination light is therefore a poor choice of illumination light color for this application.
FIGS. 4 and 5 illustrate example black-and-white images of silicon dies 4c, 4d as would be seen by a vision system of a conventional automated fluid dispensing equipment that uses blue LEDs for illumination. The silicon die 4c in FIG. 4 has a green tint. Accordingly, since the color green is not opposite the color blue on the color wheel of FIG. 1, green is not a complementary color of blue. Accordingly, blue illumination light results in good contrast between the die 4a and background 2. In FIG. 5, however, the silicon die 4d has a dark blue tint. In this case, since the color dark blue is adjacent rather than opposite the color light blue on the light color wheel, blue illumination light results in high contrast.
FIGS. 2-5 illustrate that contrast can be improved in a vision system by changing the color of the illuminating light. However, when the color of the product being processed changes frequently it is not practical to change the illumination color by replacing LEDs in the vision system. In the example of FIGS. 2-5, within one type of silicon die, the color remains consistent enough for the replacement method to work, and the blue light works well for both the green and blue die surfaces. If, however, the nitride thickness were changed slightly to produce a reddish appearing die surface, the blue light illuminating the red die would be as ineffective as the red light on the blue die as shown in FIG. 3.
In summary, while many parts have a very consistent appearance, it is possible for the optical characteristics of some parts to vary considerably during production. A slight change in the reflectivity or color of the part can significantly affect the difference in contrast between the desired feature of the part and the background. For parts subject to considerable variation, to avoid production halts or product or equipment damage, it would be desirable if the equipment could automatically adjust for varying part image conditions.
Accordingly, a need exists for a vision system which allows varying the illuminating light color and intensity. A need also exists for a method for allowing the vision system to vary the illuminating light color and intensity automatically. A need further exists for a vision system which optimizes the contrast between the desired feature of a part being processed and the background on which the part is mounted.